Research Article: Identification of qPCR reference genes suitable for normalizing gene expression in the mdx mouse model of Duchenne muscular dystrophy

Date Published: January 30, 2019

Publisher: Public Library of Science

Author(s): John C. W. Hildyard, Amber M. Finch, Dominic J. Wells, Atsushi Asakura.


The mdx mouse is the most widely-used animal model of the human disease Duchenne muscular dystrophy, and quantitative PCR analysis of gene expression in the muscles of this animal plays a key role in the study of pathogenesis and disease progression and in evaluation of potential therapeutic interventions. Normalization to appropriate stably-expressed reference genes is essential for accurate quantitative measurement, but determination of such genes is challenging: healthy and dystrophic muscles present very different transcriptional environments, further altering with disease progression and muscle use, raising the possibility that no single gene or combination of genes may be stable under all experimental comparative scenarios. Despite the pedigree of this animal model, this problem remains unaddressed. The aim of this work was therefore to comprehensively assess reference gene suitability in the muscles of healthy and dystrophic mice, identifying reference genes appropriate for specific experimental comparisons, and determining whether an essentially universally-applicable set of genes exists. Using a large sample collection comprising multiple muscles (including the tibialis anterior, diaphragm and heart muscles) taken from healthy and mdx mice at three disease-relevant ages, and a panel of sixteen candidate reference genes (FBXO38, FBXW2, MON2, ZFP91, HTATSF1, GAPDH, ACTB, 18S, CDC40, SDHA, RPL13a, CSNK2A2, AP3D1, PAK1IP1, B2M and HPRT1), we used the geNorm, BestKeeper and Normfinder algorithms to identify genes that were stable under multiple possible comparative scenarios. We reveal that no single gene is stable under all conditions, but a normalization factor derived from multiple genes (RPL13a, CSNK2A2, AP3D1 and the widely-used ACTB) appears suitable for normalizing gene expression in both healthy and dystrophic mouse muscle regardless of muscle type or animal age. We further show that other popular reference genes, including GAPDH, are markedly disease- or muscle-type correlated. This study demonstrates the importance of empirical reference gene identification, and should serve as a valuable resource for investigators wishing to study gene expression in mdx mice.

Partial Text

The X-linked muscle-wasting disease Duchenne muscular dystrophy (DMD) affects roughly 1 in 5000 new born boys [1], and is the most common fatal genetic condition diagnosed in childhood. Caused by absence or insufficiency of the muscle membrane-associated protein dystrophin, muscle fibres lacking this protein sustain damage under even normal use [2]. Repeated cycles of muscle degeneration and compensatory regeneration, alongside a steady accumulation of fibrotic scarring and fatty replacement, lead to muscle atrophy, loss of function and ultimately death. While the condition is presently incurable, DMD remains a field of active research: several different approaches aimed at dystrophin restoration are currently under investigation or entering therapeutic trials [3, 4]. Such research is aided by animal models, and multiple models of this disease exist, including mouse [5], rat [6, 7], rabbit [8], dog [9–11] and pig [12, 13]. While mouse models (particularly the mdx mouse) are the most frequently studied, each model offers discrete benefits and caveats in terms of disease severity, disease progression, cost and therapeutic tractability [14]. Regardless of species, assessment of the consequences of insufficient dystrophin -and more critically, the extent of therapeutic dystrophin restoration- utilises multiple investigative avenues, from whole animal and whole muscle physiological studies, to gross histology and immunostaining, to quantitative measures of gene expression at the protein and mRNA level.

A panel of reference genes suitable for normalizing gene expression in both healthy and dystrophic tissue (regardless of age or muscle group) would represent a valuable tool in the DMD research tool-kit, ostensibly permitting otherwise quite disparate samples (such as aged diaphragm and young tibialis anterior) to be empirically compared and evaluated, both within and between research groups. Dystrophic muscle is however host to a complex mixture of cell types, and moreover a mixture that changes with age, frequency of use and muscle type: it is by no means guaranteed that any set of genes might be appropriate for normalizing expression between dystrophic muscles, let alone between such muscles and matched healthy tissue. The study presented here attempts to address this question in a rigorous and comprehensive fashion, with a dataset using 13 genes and fully 126 samples, allowing assessment of gene suitability in multiple muscles, including the TA, diaphragm and heart, taken from mdx and strain-matched healthy mice at three disease- and research-relevant ages (6 week early disease, 10 week disease progression and 24 week established disease, with all three being commonly employed time-points for therapeutic trials). As shown, our data reveal that CSNK2A2, AP3D1, RPL13a and ACTB appear to represent a universal mouse panel (albeit not without caveats). These four genes score highly under geNorm, Normfinder and BestKeeper analysis, both within the dataset as a whole and when assessed as subsets of the data, suggesting a high level of stability. Further investigation clarified this stability: all four genes exhibit no age or muscle-group specific changes, however all four genes do show moderate (but statistically-significant) disease-associated changes (Fig 7). The changes are small and indeed below the threshold of significance when considered on an individual muscle level: only with the greater statistical power afforded by more systematic analysis do these changes reveal themselves (illustrating a major strength of our large and comprehensive dataset). Such a result might be considered problematic: use of any one of these genes as a reference would introduce a small (~20%) but nevertheless consistent disease-specific bias. The significance of our findings is that this bias can be effectively (and easily) eliminated, as the differences exhibited are almost perfect mirror images of each other: the extent to which ACTB and RPL13a are increased in dystrophic tissue is near-exactly matched by the extent to which CSNK2A2 and AP3D1 are decreased. A normalization factor prepared by geometric average of all four (or indeed one of each sign) should exhibit no net disease-associated behaviour, and would thus be entirely appropriate for normalizing gene expression data taken from both healthy and dystrophic tissue, regardless of animal age or muscle type studied.

The primary goal of this work was to establish whether a set reference genes exists for mouse muscle that remain suitable for both healthy and dystrophic samples regardless of animal age or muscle type studied: a standard panel of qPCR reference genes that remain valid under essentially all comparative scenarios would, if adopted widely, render multiple studies conducted in multiple research groups highly comparable. As shown here, despite showing small disease-associated changes, the combination of CSNK2A2, AP3D1, RPL13a and ACTB appears to fulfil this challenging remit. Moreover, we have previously shown that such a set exists for normalizing expression in healthy and dystrophic canine muscle [20], raising the enticing possibility that a similar set might be identified for use in human patient samples. We would not necessarily expect the same genes to perform well in humans: our data suggest that extrapolation between species is not straightforward. RPL13a performs well in both mouse and dogs (and thus may similarly excel in humans), but our other candidate genes show clear species-specificity. Humans are genetically closer to mice than dogs, but more similar to dogs with respect to muscle metabolism and disease presentation/severity. A human equivalent of the study presented here would be markedly more ambitious, but our data suggests such an effort might yield success.




Leave a Reply

Your email address will not be published.